Jack Dorsey’s Block has unveiled Dessa, an AI voice model trained “from scratch on nothing but human conversation” — and the result talks to you like an awkward new friend, with all the uncanny back-and-forth that implies.
Dessa AI voice model (Credits: Cath Virginia / The Verge)
Block’s Dessa team, led by Ragavan Thurairatnam, shared demo conversations to show off the model. In one, Dessa plays a guessing game over voice — and accidentally calls the human speaker “Ronaldo” before recovering. On the Dessa website there are more: a back-and-forth about Christopher Nolan films that veers onto an unexpected tangent about Bryan Singer’s Superman movies, a tearful-sounding reaction to a surprise Paris trip, and a voice conversation with OpenAI’s GPT-Live-1 where Dessa sounds notably more natural.
Dessa: True Speech-to-Speech AI That Runs Locally
Unlike today’s voice assistants that flatten speech into text first, Dessa is a true speech-to-speech model — “everything it knows came from human speech.” The team says it can run locally on a MacBook Pro M4 Max or better.
The motivation, per the Dessa team, is threefold: curiosity, better understanding between humans and AI, and safety. They frame it simply: babies learn to talk without ever seeing a word written down, so they wanted to see if a model could too. And on safety: a voice AI that cannot hear the emotion in a voice is dangerous — “imagine being in serious pain while a voice AI has no idea and just keeps going.” Dessa is a step toward fixing that.
When Will Dessa Be Available?
There is no public release yet. Block says it will give “a small group of AI safety researchers” early access to the model first. Whether Dessa grows into a consumer product or stays a research experiment, it is one of the most interesting bets yet on voice-first AI.
Also Read: How OpenAI Scaled Storage to Serve Over 1 Billion ChatGPT Users Across 40 Regions
Source: The Verge — read the full story there.